An Integrated Framework for Explainable, Fair, and Observable Hospital Readmission Prediction: Development and Validation on MIMIC-IV

Abstract

Objective: To propose and retrospectively validate an integrated framework that simultaneously addresses three barriers to clinical translation of readmission prediction: lack of explainability, absence of deployment reliability infrastructure, and inadequate demographic fairness evaluation. Materials and Methods: A cohort of 415,231 adult admissions from the MIMIC-IV clinical database (30-day readmission prevalence 18.0%) was split chronologically 70/15/15. Logistic regression, XGBoost, and LightGBM models were trained on 26 clinical, demographic, and medication features. SHAP TreeExplainer provided per-patient feature attributions. Fairness was evaluated across 16 subgroups spanning race/ethnicity, age, gender, and insurance type using AUC-ROC, false negative rate (FNR), and positive predictive value (PPV). Calibration was assessed via Brier scores and calibration curves. A deployment-ready observability architecture was specified using Prometheus, Grafana, and Azure Kubernetes Service. Results: XGBoost achieved AUC-ROC 0.696 (95% CI: 0.691-0.701), outperforming or matching the LACE clinical baseline (AUC 0.60-0.68). LightGBM achieved the best calibration (Brier score 0.146). Prior admissions in the preceding 12 months were the dominant SHAP predictor (mean |phi| = 0.085). All 16 demographic subgroups met equity thresholds (ΔAUC ≤ 0.05, ΔFNR ≤ 0.10) without post-processing. Discussion: The framework jointly addresses explainability, fairness, and deployment reliability - requirements not previously integrated in published readmission prediction systems. Conclusion: This integrated framework delivers competitive discriminative performance, clinically actionable per-patient explanations, and strong demographic equity simultaneously. All code is publicly available at https://github.com/Tomisin92/readmission-prediction